Proof & method
The experiment is this website.
It would be a bit rich to sell a service that improves websites automatically, using a website we polished by hand. So this one is the guinea pig: same system, same rules, same public record — and we posted the starting numbers before doing any of the work.
Baseline · 2026-06-04 → 2026-09-02
Source: Google's own search figures for this website
- Times shown
- 18,001
- Visits
- 26
- Click rate
- 0.144%
- Typical spot
- 39.9
people who saw us in Google
people who actually clicked
the gap, in one number
about page four of Google
What the baseline says
In the 90 days to 2026-09-02, this website appeared in Google 18,001 times and got 26 visits from it. That is roughly one visit for every seven hundred times we were shown, sitting on average around page four.
Sitting around page four means Google knows the site exists and thinks it is roughly relevant, but nobody scrolls that far. Those eighteen thousand appearances are real people looking for something we could have helped with. That is a better starting point than no interest at all, because the demand is already there — it just is not being met.
The demand that is already arriving
These are the searches where this site gets shown but almost never clicked. Look at what they have in common: they are all describing the exact problem Autopilot exists to fix.
| What people searched for | Times shown | Visits | Click rate | Typical spot |
|---|---|---|---|---|
| web maintenance services | 1,263 | 0 | 0% | 36.4 |
| webmaster services | 1,032 | 1 | 0.1% | 41 |
| web maintenance | 724 | 0 | 0% | 50.2 |
| web maintenance and support | 590 | 0 | 0% | 55.1 |
| website maintenance and support | 525 | 0 | 0% | 64.8 |
| web maintenance service | 518 | 0 | 0% | 37.1 |
Source: Google's own search figures for this website, 2026-06-04 to 2026-09-02.
Speed, published
How fast this site actually is.
A service that sells page speed should show its own. These are updated automatically, and they include the number we cannot report yet.
Code sent to your browser
4.6KB
A typical site ships 300KB to 2MB
Heaviest page
36KB
/
Fonts
29KB
One file, served from this site
Other companies' code loading first
0
Nothing blocks the page appearing
The number we cannot show you yet
Everything above is a test, not a measurement. The figures Google actually grades you on come from real visitors on real phones, and Google only reports them once a site has enough traffic. This one does not yet — which is exactly what you would expect from a site that got 26 visits in 90 days. When there is enough traffic to report, those numbers appear here, whatever they say.
Last measured 6 September 2026
Method
How we decide something actually worked
01
Post the starting numbers first
Written down before a single change is made. That removes the oldest trick in the trade: quietly choosing a flattering start date once you already know how it turned out.
02
Leave some pages alone
Similar pages are split at random into two groups, and one group is deliberately never improved. That genuinely costs something — and it is what buys you a number you can trust.
03
Agree the timescale up front
28 days, decided in advance, with the page left untouched throughout. Anything shorter is measuring noise. Anything decided afterwards is measuring wishful thinking.
04
Report the difference, losses and all
How the improved pages moved, minus how the untouched ones moved. Anything that lost gets put back, and published just as visibly as anything that won.
What we cannot honestly claim yet
This started on 2026-09-05. Search results move slowly — a change takes weeks to settle, and a fair comparison needs a full month before it means anything. Anyone showing you dramatic results a few days after starting is showing you random movement and calling it progress.
So there is exactly one honest claim on this page today: here is where we started, here is how we will measure it, and here is everything we have done. The next set of numbers goes up when the first month is done — whether or not they make us look good.
Want the same thing run on your site?
It starts the same way every time: your starting numbers written down, a group of pages left alone to compare against, and an agreed timescale.